From Rodney Smith
Noel Sharkey’s rejection of the computational theory of mind – the theoretical foundation of artificial intelligence – is a breath of fresh air (29 August, p 28). As Sharkey rightly says, the intelligence is in the human who creates the program, not in the program itself.
But in making this point, Sharkey goes further than denying that a program could be intrinsically intelligent: he also asserts that the machine that executes the program could not be intrinsically intelligent. This conceptual jump from computation to machinery reveals an assumption: namely that computers are limited to performing computations. It ignores the possibility that if computers could do more than compute, they might be able to perform the non-computational operations that are a necessary part of intelligence and mind.
This idea is not idle speculation. My own and other research indicates that computers are capable of performing various types of non-computational operations, including a type of associative processing in which inputs are stored according to their association to each other. This gradually builds an inner structure not prescribed by a programmer. Which of the inputs are associated with one another to form this structure is not specified by a human; characteristics of the input stream itself determine the structure of the system.
This implies that it is possible for a computer to be intelligent, even if no program can be. The fault in AI is the limited view it takes of the computer as a computing machine.
Advertisement
In associative processing there are three elements: the program, the machine, and the inner structure built from input symbols where the input stream itself – not a human designer – determines the shape of the output. A tantalising prospect is that such a structure could be a mind.
Wellington, New Zealand
